A Label‐Free Hyperspectral Imaging Device for Ex Vivo Characterization and Grading of Meningioma Tissues

Pietro Ricci, Camilla Bonaudo, Ivan Ezhov, Anam Toaha, Dorotea Nardini, Manuel Camelia, Federica Lucidi, Filippo Nozzoli, Tim Mach, Ilias Tachtsidis, Daniel Rueckert, Alessandro Della Puppa, Luca Giannoni, Francesco S. PavoneJournal of Biophotonics 2025 · DOI

TL;DR

This paper introduces HyperProbe1.1 (HP1.1), a novel, label-free hyperspectral imaging system designed for rapid, ex vivo characterization of brain tumor tissues. By analyzing the biochemical and metabolic signatures of meningioma samples, HP1.1 demonstrates the potential to provide real-time grading information, complementing traditional histopathology and aiding intraoperative decision-making.

What Problem Does This Paper Address?

Currently, definitive tumor diagnosis and grading rely on histopathology, a process that involves tissue fixation, processing, and staining, taking several days. This delay can critically impact postoperative treatment planning, especially for aggressive tumors. While intraoperative tools are needed to guide surgical decisions in real-time, existing methods often lack the ability to provide detailed biochemical information or only sample small areas. Meningiomas, the most common primary intracranial tumors, present a significant clinical challenge due to their varied grades and recurrence risks, making rapid and accurate grading crucial for patient management. This paper aims to bridge this gap by developing and validating a hyperspectral imaging system for fast, label-free biochemical assessment of meningiomas.

Methods

The core of this work is the HyperProbe1.1 (HP1.1) system, a compact and transportable hyperspectral imaging (HSI) device. HP1.1 utilizes a broadband plasma light source (250-2500 nm) coupled with a flexible wavelength selector to acquire spectral data across the visible and near-infrared range (385-1015 nm) with user-defined bandwidths. Light is delivered to the sample via an optical fiber and reflected light is collected by a macro-objective and focused onto a scientific CMOS camera.

Key methodological aspects include:

  • Hypercube Generation: The system captures a series of 2D reflectance images at discrete wavelengths, forming a hyperspectral "hypercube" where each pixel contains a full spectral profile. This allows for pixel-wise spectral analysis.
  • Data Calibration and Processing: Standard procedures for dark subtraction and normalization to a white reference are employed to obtain reflectance spectra. Absorption and scattering are modeled using a logarithmic transformation.
  • Spectral Unmixing: A modified Beer-Lambert Law (MBLL) approach is used to decompose the reflectance spectra into constituent chromophores. This allows for the quantification of endogenous biomarkers like water, lipids, oxy- and deoxyhemoglobin, and cytochrome c oxidase (CCO). The analysis focuses on the 500-900 nm range for optimal tissue penetration and biomarker detection.
  • RGB Reconstruction: For visualization, RGB composite images are generated from specific spectral bands, highlighting spectral contrast and spatial patterns.

Datasets

The study utilized meningioma tissue samples obtained from 21 patients undergoing surgery at the University Hospital of Careggi in Florence. These fresh, ex vivo samples were imaged within 1 hour of surgical removal. For each specimen, multiple fields of view (FOVs) were acquired, totaling 43 FOVs from the 21 patients. The histopathological grading (WHO grades 1, 2, and 3) was performed by the Pathology Unit of the same hospital, serving as the ground truth for comparison. No new benchmark datasets were introduced; rather, existing clinical and histopathological data were used for validation.

Key Insights and Ideas

  • Label-Free Biochemical Fingerprinting: HSI, through spectral unmixing, can extract detailed biochemical information (lipids, hemoglobin oxygenation, CCO activity) from fresh tissue without the need for stains or labels.
  • Correlation with Tumor Grade: Distinct molecular patterns, particularly in lipid content, hemoglobin oxygenation, and CCO activity, were observed to correlate with meningioma histopathological grades.
  • Intra-Tumor Heterogeneity: The system effectively captures spatial variations within a single tumor, revealing differences in metabolic activity and vascularization that might be missed by single-point probes.
  • Potential for Intraoperative Guidance: The rapid acquisition and analysis capabilities of HP1.1 suggest its potential integration into surgical workflows to provide real-time feedback, aiding in tumor localization, margin assessment, and potentially guiding surgical strategy.
  • Complementary to Histopathology: While histopathology remains the gold standard, HSI offers a complementary, faster, and label-free approach that can provide crucial biochemical context.

Key Results

  • The HP1.1 system successfully acquired hyperspectral data from fresh meningioma biopsies, enabling pixel-wise spectral analysis.
  • Spectral unmixing allowed for the quantification of endogenous biomarkers, revealing differences in their distribution across different WHO grades of meningiomas.
  • Grade 1 meningiomas generally showed lower lipid content and higher oxidized CCO, suggesting a less aggressive, well-perfused tissue.
  • Grade 3 (anaplastic) meningiomas exhibited higher lipid content and more heterogeneous DiffCCO distribution, indicative of increased vascularization, metabolic activity, and potentially compromised oxygen delivery.
  • Qualitative trends suggested that HSI can differentiate between tumor grades based on spectral signatures, with grade-related clustering observed in parameter plots (e.g., DiffCCO vs. lipid content).
  • The system demonstrated good spectral agreement between ground-truth attenuation spectra and model-inferred attenuation, validating the spectral unmixing approach.

Limitations and Future Work

  • Limited Dataset Size: The study acknowledges that the small and unbalanced dataset (especially for grade 3 cases) prevents robust statistical inference and classifier development.
  • FOV-Level Analysis: While multiple FOVs were acquired for larger specimens, they were not individually coregistered to distinct H&E sections, meaning FOV-level differences might reflect genuine intratumoral heterogeneity rather than grading inaccuracies.
  • Ex Vivo Application: The current study focuses on ex vivo analysis. The authors propose future work to validate these findings in vivo and explore their use in intraoperative settings.
  • Preclinical Validation: Further preclinical research in animal models is suggested to evaluate how systemic physiological elements affect hyperspectral measurements and to aid in translational research.
  • Broader Applications: The authors envision extending this technology to other brain tumors, including gliomas and metastases, to improve intraoperative discrimination and inform surgical strategies.
  • Integration with Surgical Workflows: Future efforts will focus on integrating HP1.1 into surgical workflows and potentially combining it with neuronavigation systems.

Summary updated

Abstract

Histopathology remains the gold standard for definitive tumor diagnosis after surgical resection; however, its lengthy processing time can delay critical postoperative care. Hyperspectral imaging (HSI) is emerging as a promising label-free technique for rapid biochemical tissue assessment. Here, we present HyperProbe1.1 (HP1.1), an HSI system designed for noninvasive analysis of fresh brain tumor biopsies. In this proof-of-concept study, we applied the HP1.1 system to freshly excised meningioma specimens-the most common primary intracranial tumors. The platform enabled rapid, label-free mapping of metabolic activity and vascular heterogeneity, while spectral unmixing further allowed the quantification of endogenous biomarkers such as cytochrome c oxidase (CCO), hemoglobin derivatives, and lipids, revealing molecular patterns consistent with histopathological tumor grading according to the 2021 WHO classification. These results highlight the feasibility of HSI for rapid biochemical tissue assessment and its potential integration into intraoperative decision-making.
BibTeX
@article{ricci2025labelfree,
  author = {Pietro Ricci and Camilla Bonaudo and Ivan Ezhov and Anam Toaha and Dorotea Nardini and Manuel Camelia and Federica Lucidi and Filippo Nozzoli and Tim Mach and Ilias Tachtsidis and Daniel Rueckert and Alessandro Della Puppa and Luca Giannoni and Francesco S. Pavone},
  title = {{A Label‐Free Hyperspectral Imaging Device for Ex Vivo Characterization and Grading of Meningioma Tissues}},
  journal = {Journal of Biophotonics},
  year = {2025},
  doi = {10.1002/jbio.202500374},
}
Figure